Oxagile AI · Case studies
Work that shipped
13 engagements, each with the results it reported. The ones with no AI in the product are here because the same delivery standard shipped them, on the pipelines our AI work runs on. One is our own tooling, labelled as such.
OTT & Media
Rights, metadata, telemetry
AI in production
Highlights cut while the match is still running
72% reduction in video processing time
A model that watches the live feed, finds the moments worth replaying, and cuts the compilation before the second half.
AI in production
Point a phone at the court, get the player
2-3 seconds to recognise players during a game
Player recognition from a phone camera during a live game, with statistics on the screen before the play is over.
Partly AI
One AI search app across 7 TV and mobile platforms
3 months to deliver
An AI search agent for films and series, with ratings, reviews, age guidance and trailers in one place, launching each title in the service that carries it. Built as an MVP for fundraising.
Partly AI
Transcoding errors forecast a month ahead
7,000+ SD and HD channels monitored
A monitoring system for more than 7,000 channels that reads early warning signs in the broadcast metrics and forecasts transcoding errors up to a month before viewers would see them.
Engineering, no AI
A pay-TV platform modernised while it stayed live
Component re-renders cut by 50%
The front end of a white-label video platform, revamped feature by feature inside the vendor’s own team: new modules, a faster interface and a quieter API.
Healthcare
Records, imaging, triage
FinTech
Payments, lending, risk
Engineering, no AI
Payment routing for MENA, LATAM and Southeast Asia
80% faster processor integration, from months to weeks
A single payment provider stopped working once the client sold in new regions. An orchestration layer now routes each transaction by rule, and reroutes it when a provider fails.
Engineering, no AI
Card payments moved to a new provider, live
1M+ users’ card payments migrated with zero downtime
Live card payments moved off one gateway onto another, with 3-D Secure added and payment tokens carried across, while the business kept taking orders.
Engineering, no AI
Treasury software built to Sarbanes-Oxley controls
84 months of continuous delivery
An online cash management system for enterprise treasuries, covering payments, FX, cash pooling and netting, with Sarbanes-Oxley controls built into the product.
Retail
Search, catalogue, demand
AI in production
Millions of reviews, asked a question in plain English
15 seconds or less to answer a question
Topic modelling over millions of customer reviews, with a conversational agent on top so the answer takes seconds rather than an analyst.
AI in production
Printed promotions, cropped and read by machine
99.3% cropping accuracy
Computer vision crops each advertisement off the page, reads printed and handwritten text, and classifies the offer, at a volume no manual team could match.
AI in production
Unwanted reviews flagged before a person reads them
80% accuracy predicting unwanted reviews
A fine-tuned BERT model reads incoming reviews for spam, off-topic and negative content, including irony and sarcasm, and tells the team which ones to remove.
Our own tooling
Internal, labelled as such
Next step
Yours would be number 14.
Tell us where it breaks. You get a written answer on what breaks first and what it costs.